A New Efficient Conformation Search Method for ab initio Protein Folding

Jae-Min Shin1, Dai Sig Im2, Byungkook Lee
1jms@idrtech.com, IDR Tech. Inc.; 2idscom@idrtech.com, IDR Tech. Inc.

We devised a new conformation search method, called Window Growth Evolutionary Algorithm (WGEA), for ab initio protein structure prediction. It searches conformations using many small windows along the sequence of the protein, then the window size is gradually increased until it reaches the length of the whole protein. Conformations are generated and tested by Metropolis Monte Carlo procedure. Many windows are selected and mutated in series before the window size is increased by one residue. Multiple chains are subjected to the same procedure in parallel and best chains are selected among all offsprings of all parents. Interactions between residues beyond the selected window are ignored and the fitness of a chain is determined by the energy of the selected window only. This procedure was tested on off-lattice, discrete state models of a set of small proteins. When coordinate root mean square (cRMS) deviation from the experimental structure is used as the energy function, WGEA folds these proteins to an average accuracy of 1.5 for a-helical proteins, 2 for the a/b class proteins, and 2.8 for the b-class proteins. In contrast, a conventional Monte-Carlo method, which is like WGEA but which does not have the window growth feature, give structures with RMS values greater than 4 for the same set of proteins. When distance RMS (dRMS) deviation with cut-off distance was used as the potential, WGEA refolds the same proteins to 2-3 cRMS.